Altitude is very incorrect

Ports in my VA3 are clean and I calibrate the altimeter before every run. I consequently get wrong readings. Attaching another example - I started and ended my run at the same place and device shows about 50m of difference. I'm trying to get help from Garmin support on this. So far I got two answers from them. In the first one they described the limitations of using GPS to measure altitude. When I mentioned that VA3 has also barometer they described the limitations of using barometer. It looks that they just give generic answers and don't even bother to check what device the complain is about.ciq.forums.garmin.com/.../1352165.png

Can't ride my mount for a while now for a serious mechanical issue to be fixed, then I took the watch (VA3) off more frequently and left it unused for hours, sometimes. This made me notice something maybe obvious but probably interesting to this discussion. I have elevation always showing up in the watch face, and also have the myABC widget installed.
The elevation strongly depend on what the internal temperature sensor reads. I use the Temperature Widget from Connect IQ Store to read what the internal thermometer measures.
The temperature changes between 31°C and 29°C approx. (88°F- 84°F approx.) depending on whether I wear the watch on the wrist or leave it rest on a table near an open window. Correspondingly, I observe the elevation changes of the order of 50-100m (15-30ft), the higher the temperature the lower the elevation.
I know how a barometric altimeter works in principle but have any knowledge on how the specific device in the VA3 works in practice, then I don't have any clue on how this relate to the details of the altimeter implementation, but this phenomenology makes me suspect that relying on the internal watch temperature sensor changes might may be the culprit. This would explain why the elevation measured during my activities seems to me to be reasonable at the beginning, after calibration and for few minutes, and completely messed up after a few kilometers, when probably skin temperature raises and temperature readings completely mess up the barometric formula implementation. I suspect (but I can't verify it at the moment, waiting for my mount to get serviced), that using the VA3 mounted in the handlebar, instead of wearing it, would give more reasonable elevation readings (but of course this is not a workaround for a device which has been designed to be a watch). I'll try it as soon as possible, but maybe someone might try it as well.

I hope Garmin developer team is reading this thread, and then, again, pronounce something concrete about this annoying issue.

I used to have a decent temperature chamber at my disposal, but we returned it since it was a rental. It would have been fun to run this through and see what kind of correlation between temperature and elevation there is.

I would suspect the temperature is being used to help compensate for error in the barometer sensor, but the math is wrong somehow. Two degrees change of wrist temperature should not result in 30ft of change unless the barometer sensor is really that touchy and poor. If the compensation was truly correct, the temperature change would not create a shift in elevation at all. Sensors I have used in the past had compensation built-in so it wasn't even part of the equation.

Guys, don't expect major improvement here. According to Garmin support the chart I attached to my previous post is acceptable...

A quick test using my wrist, a desktop, and a fridge. Altitude was set manual, automatic off. When I set the altitude to my location, the sea-level barometer was exact with a nearby weather station, so that is all excellent.

30.6C = 593ft
23C = 620ft
18C = 644ft (I didn't get to let it chill for long)

So it seems elevation does increase a little while it gets colder, but not enough to care: 4ft/C with this test.

In order to get hundreds of feet of error must be from a different source. Likely GPS is getting thrown into the mix.

Next test, when I have time, is to try the same thing with the GPS on, though it will be hard to get a cold reading in a fridge and still see the GPS.

I attach two pics to show better what I meant.
First one watch on wrist, temperature stable around 32.5°C.
Second one, watch on the desk, temp around 26°C. Elevation difference around 50m.
The Elevation graph shows many repetitions of the experiments during relax time in my living room, in very stable weather conditions, with peaks at around 55m, which is approx the elevation at my place, and dips at around 10m anytime I wear the watch on.
The baro graph shows almost constant sea level pressure, despite spikes and steps, the overall excursion is fractions of a millibar.

Also the same issues, running on sea level today but if I must believe the VA3 I have done a run in the Alps, today almost 700 meters above SL.

Same thing. Today I'd did a 60km ride along the coast - a very flat ride. The altitude graph showed a lot of small oscillations around sea level. For some reason my Garmin instinct sums up the noise on the altitude and at the end of the ride tells me I have climbed and descended 3000m!! Interestingly, when strava analysed the ride from Garmin data it give a reasonable cumulative ascent.

I no longer own a VA3 as of last December, but my wife still has one and she has the same issues. Here's the bottom line. VA3 owner's need to bombard Garmin with phone calls, emails and even a petition to bring back manual calibration of both the altimeter and the barometer. The altimeter and barometer are sensors that require calibration. Whatever auto calibration Garmin is now using is hit or miss. Mainly miss. If and when this is restored, there are three factors people need to keep in mind prior to doing an activity. 1. Calibrate the altimeter to a known elevation. This can be done using the internet for where you are located, or there are a couple of IQ apps that work well. 2. Calibrate the barometer to the correct MSL pressure prior to starting an activity. Again, this info can be obtained through the internet or a widget that displays current pressure at your location. 3. And I don't know why Garmin doesn't explain this in more detail, but try and have the watch at the approximate temperature you'll be doing your activity. The barometer is temperature sensitive. Once it starts to warm, the barometer will rise a little and elevation will drop. This may be what paulgear saw in his graph in post #46 where it appeared the hills were getting smaller. This is easier said than done, but going from an air conditioned building to a warm outside temperature (or vise versa) will cause these "odd" elevation readings. Don't give up that manual calibration is done and gone. If nothing is said, nothing will be done. The forum is a great place to pass the word on to fellow users, but it's contacting Garmin that will be the only way to get things changed.

Notice how the hill gets lower on each successive run (I wish!). GPS was enabled, so any blocked pressure sensors shouldn't make any difference. Is this just a fault?

It's not a fault and it's actually performing as it should. If you look at your graph you'll notice you peaks and valley's are about the same. This means your altimeter is working properly displaying your elevation gain and loss. The reason your total elevation decreased even though you ran the same hills is because your barometer was rising, which in turn decreases elevation. This may have been caused by an actual increase in barometric pressure, temperature (actual or body) rising, or a combination. This is why I say it's important to calibrate the altimeter and the barometer prior to an activity with the watch temperature as close to the actual temperature the activity will be performed. If Garmin restores manual calibration again, and you pay attention to these things before, during, and after a workout, you'll know why you have this readings and they'll make perfect sense.